Treatment of adults with autoimmune thrombocytopenic purpura (AITP) is based more on individual experience than on results of controlled studies. We compared intravenous immunoglobulin with high-dose methylprednisolone in untreated adults with severe AITP and assessed efficacy of subsequent oral steroids compared with placebo. Primary outcome was number of days with platelet count greater than 50×109/L within the first 21 days. We did a randomised multicentre trial based on a 2×2 design. 122 adults with severe AITP (platelet count ⩽ 20×109/L) were randomly assigned to receive either intravenous immunoglobulin or high-dose methyl-prednisolone on days 1–3 (randomisation A), and then to receive either oral prednisone or placebo (randomisation B) on days 4–21. Analysis was by intention to treat. Six patients were excluded from the analysis. The number of days on which platelet counts were above 50×109/L was 18 in 56 patients receiving intravenous immunoglobulin and 14 in 60 receiving high-dose methylprednisolone (p=0·02). Percentage of patients who had platelet counts over 50×109/L on days 2 and 5 was 7% and 79%, respectively, in the intravenous immunoglobulin group compared with 2% and 60%, respectively, in the high-dose methylprednisolone group (p=0·04). During the second treatment period, prednisone was more effective than placebo for all short-term endpoints. Patients who received intravenous immunoglobulin and prednisone had platelet count greater than 50×109/L for 18·5 days (p=0·005), and those treated with high-dose methylprednisolone and prednisone had this count for 17·5 days. Intravenous immunoglobulin and oral prednisone seems to be more effective than high-dose methylprednisolone and oral prednisone in adults with severe AITP, although the latter treatment is effective and well tolerated.